Challenges
Access Data
The best way to access and download the data is via NOAA’s Open Data Dissemination bucket for the DCLDE 2027 workshop:
Important Dates
Tentative submission deadline for workshops & abstracts : 1 March 2027
Tentative notification of acceptance deadline: 5 April 2027
Confirmed conference dates: 1-6 August 2027
Detection and Classification (DC) Challenge
Killer whales (Orcinus orca) occur in multiple populations around the world, each with distinct behavior, diet, and conservation status. In the Northeast Pacific, these include three ecotypes: fish-eating Resident populations (Southern Resident, Northern Resident, and Southern Alaskan Resident), mammal-eating West Coast Transient (Bigg’s) population, and the Offshore population that specializes on elasmobranchs. Vocal dialects and acoustic behaviors differ within and among populations.
In the Salish Sea, a key monitoring priority is distinguishing endangered Southern Resident killer whales (SRKW) from co-occurring and more numerous Transient (Bigg’s) killer whales (TKW), as well as from acoustically similar species such as humpback whales (Megaptera novaeangliae) and Pacific white-sided dolphins (Aethalodelphis obliquidens). The DCLDE challenge centers on two main tasks: detecting killer whales and classifying populations.
Challenge Lead:
Kaitlin Palmer (kpalmer@coa.edu)
Participant Tasks:
(DC-1) Detect killer whale vocalizations in diverse and noisy underwater recordings while minimizing confusion with other acoustically active species.
(DC-2) Classify detected calls to population, with emphasis on distinguishing SRKW and TKW.
Dataset and Evaluation
The training dataset includes annotated recordings from drifting recorders and fixed hydrophone deployments across multiple sites, covering the three ecotypes found in the Northeast Pacific. Evaluation (dataset to be released in 2027) will consider detection accuracy, population classification performance, and generalization to novel datasets.
Link to Training Dataset, DOI: 10.25921/15ey-mh50
Dataset Description: https://www.nature.com/articles/s41597-025-05281-5 (To ensure proper attribution, please cite this paper in any publications, presentations, or derivative datasets that utilize these data).In the Salish Sea, a key monitoring priority is distinguishing endangered Southern Resident killer whales (SRKW) from co-occurring and more numerous Transient (Bigg’s) killer whales (TKW), as well as from acoustically similar species such as humpback whales (Megaptera novaeangliae) and Pacific white-sided dolphins (Aethalodelphis obliquidens). The DCLDE challenge centers on two main tasks: detecting killer whales and classifying populations.
Localization (L) Challenge
Distributed Acoustic Sensing (DAS) converts existing fiber-optic cables, such as those used by the telecommunications industry, into dense listening arrays of strain sensors that extend for tens of kilometers. These sensors can detect low-frequency whale vocalizations and are installed from shore, providing real-time data access.
The spatial dimension of DAS inherently offers positional information; visualizing the recorded spatio-temporal data matrices allows for determining the acoustic source position along the fiber. However, resolving left-right ambiguity can be difficult since fibers are often laid in straight lines, posing similar localization challenges to fixed linear arrays. This challenge will utilize DAS data recorded at the OOI cabled observatory off Pacific City, Oregon. Data were collected simultaneously on two “parallel” fibers with curved layouts, enabling the resolution of the left-right ambiguity and providing additional information such as the depth of the vocalizing animal.
Challenge Lead:
Léa Bouffaut (lea.bouffaut@cornell.edu)
Participant Tasks:
(L-1) Localize fin whale vocalizations using DAS data collected on a single fiber.
(L-2) Localize fin whale vocalizations in 3D using DAS data collected on two fibers.
Dataset and Evaluation
The training dataset contains labeled DAS recordings from two parallel DAS systems, with arrival time picks along each fiber, associated UTC timestamps, and ground-truth localization outputs (Lat/Lon and UTM coordinates).
Evaluation will be conducted on 10 minutes of labeled test data (without position information) released in 2027 before the workshop. Participants must submit predictions for each labeled call, including date (UTC), position coordinates (UTM or decimal Lat/Lon) with uncertainty estimates—XY coordinates for Task L-1, and XYZ coordinates for Task L-2.
Link to DAS Dataset, DOI: 10.25921/v2vh-8w16
Additional DAS data (same experiment) can be downloaded here: NSF_OOI_DAS.
Dataset Description: Goestchel et al. (accepted): Automated Association of Fin Whale Calls for Localization Using Distributed Acoustic Sensing. In: Journal of the Acoustical Society of America. (To ensure proper attribution, please cite this paper once available in any publications, presentations, or derivative datasets that utilize these data).
Density Estimation (DE) Challenge
For this challenge, we provide a combined dataset that includes both fixed PAM sensors and visual aerial data for North Atlantic right whales (Eubalaena glacialis) in Cape Cod Bay, MA, USA, offering multiple analytical approaches. The main goal is to estimate the density or abundance of North Atlantic right whales in the study area using acoustic data and other auxiliary data as needed. This dataset was analyzed in Garcia & Tolkova et al. (2025; link below), and various data products are being made available for this challenge.
Important Note: The visual aerial data are currently limited to aggregated daily sighting counts. However, it is possible to obtain location information for individual whales. This will require a data-sharing agreement with the Center for Coastal Studies, MA, USA. Rather than setting up multiple agreements, Danielle will be coordinating this. So please do reach out to her should you be interested in the more detailed visual dataset.
Challenge Lead:
Danielle Harris (dh17@st-andrews.ac.uk)
Participant Tasks:
(DE-1) Estimate the density (or abundance) of North Atlantic right whales in Cape Cod Bay, MA, USA, using multiple data streams.
Dataset and Evaluation
Detailed comments:
Some initial time-difference-of-arrival analyses have been performed to match right whale calls across the array. These results can be provided, or analysts may choose to conduct their own signal processing.
Detector performance was quantified in Garcia & Tolkova et al. (2025), but false positives were not removed from the dataset.
Data on call production rate, which is often a key parameter needed for density and abundance estimation analyses, is not explicitly available here. Analysts may need to consult the literature for information on behavioral parameters and should acknowledge relevant assumptions and limitations when using such data.
For the visual aerial data, there is potential to estimate aspects of visual detectability if needed.
Link to Acoustic Dataset, DOI: 10.25921/ab3a-c842
Link to Acoustic Detections & Aerial Survey Data, DOI: 10.25921/x68p-cj94
Dataset Description: https://www.int-res.com/abstracts/esr/v56/esr01384 (To ensure proper attribution, please cite this paper in any publications, presentations, or derivative datasets that utilize these data).